TasteMiner: Mining partial tastes for neighbor-based collaborative filtering

摘要

Neighbor-based collaborative filtering is one of the most practical recommendation approaches that is renowned because of its simplicity and explanation. However, the big limitation is its high computational complexity. It is demonstrated that clustering-based algorithms, that restrict the neighborhood space, speed up the recommendation process at the price of lower accuracy. We propose a new algorithm, called TasteMiner that efficiently learns partial users taste to restrict the neighborhood space. We frame TasteMiner as a method for neighborhood collaborative filtering, and show its effectiveness compared to previous algorithms